Most firms buy legal tech before they can point at the step it fixes. We place a Navigator inside one firm at a time — an embedded operations engineer — to map how the work actually moves and put a written decision behind every place AI does or does not belong.
That reaction is not technophobia. It is the most lucid response in the room. Nobody in that meeting, including the vendor, could tell you which step of your matter lifecycle the product plugs into, because nobody had mapped it.
The org chart shows who reports to whom. It does not show how a matter moves from intake to close, where it waits, or who touches it twice. Without that, every tool decision is a guess.
If the only complete copy of a procedure is a senior paralegal's memory, it is not a process. It is a dependency, and it walks out the door at five o'clock.
Firms drift into using AI on work they never authorised it for, because the boundary was never documented. That is not a capability gap. It is a governance decision that nobody made on purpose.
Two published frameworks, applied in order. The map has to exist before the decision can be made, which is the step almost every AI adoption effort skips.
A production map of how a matter actually moves through the firm, built around the work rather than the hierarchy. It shows the handoffs, the queues, and the steps that have quietly stopped earning their place.
Read the white paperEvery mapped step gets classified: Automate where the rules are settled, Augment where a human keeps judgment, Abstain where AI has no business at all. The reasoning is recorded at the moment of classification, not reconstructed afterwards.
Read the white paperWhat classification produces
| Class | What it means | The test |
|---|---|---|
| Automate | The step runs without a human in the loop. | The rules are fully settled and a wrong output is caught downstream before it reaches a client. |
| Augment | AI prepares; a person decides and signs. | Judgment is required, or the output is client-facing. This is the default when the call is close. |
| Abstain | No AI, by written decision. | The step is the practice of law, or the confidentiality and privilege exposure is not worth the gain. |
Most operations consulting is a visit: interview eight people, write down what they said, leave a document. The document is usually right and almost never implemented, because the person who understood it is gone. We place a Navigator inside one firm for six to twelve months instead, working in your matters and your systems until the changes hold.
The engagement is front-loaded, then tapers into a retainer
Heaviest
Months 1–3
Production mapped across your main matter types, every step classified, the written policy drafted, and a costed list of what to fix first.
Heavy
Months 4–7
The roadmap gets executed rather than handed over: procedures, reporting, and automations built in your own systems.
Tapering
Months 8–12
Staff trained, owners assigned to every procedure, and the consultant deliberately stops being load-bearing.
Retainer
Ongoing
Quarterly review, reporting upkeep, and policy re-review when bar guidance moves.
Law firms carry privilege obligations that most consultants have never had to think about. These are commitments, not aspirations, and they are written into the engagement agreement.
Everything built during the engagement belongs to the firm, in the firm's own accounts. We keep only the general methods we arrived with. If the relationship ends, nothing stops working.
Each engagement runs in its own tenancy with structural separation between firms. No shared workspace, no cross-client data, no client material used to train anything.
We do not resell software and we take no referral fees. When the recommendation is to keep what you have, or to buy nothing, that recommendation costs us nothing to make.
The policy output is built against ABA Formal Opinion 512 and current state guidance, with review triggers so it does not go stale when the rules move.
Both frameworks are out in full, at length, with the reasoning intact. You can decide whether we think clearly about your problem without spending anything to find out.
Why hierarchy charts hide the way work actually moves, what railroad managers worked out in the 1850s, and how to map production instead of position.
Read the white paper Article twoA decision model for AI in a law firm, and why the third option is the one that makes the other two defensible.
Read the white paperWe do not sell software and we take no vendor commissions. The engagement produces a map of how your firm actually works and a written decision about which steps to automate, which to augment, and which to keep entirely human. Tools get selected after that map exists, if they are needed at all.
Firms roughly in the four to ten million dollar band, typically one attorney-owner with up to about five non-owner attorneys. Large enough that process failures cost real money, small enough that the owner can still change how the firm works by deciding to.
The firm keeps its data and its instance of everything built during the engagement. Hopper Oak keeps only the general methods it brought with it. That split is written into the agreement rather than left to good intentions.
Two, with staggered start dates. This is an embedded consultant, not a portfolio of accounts, so availability is genuinely limited and worth asking about early.
A short call, no deck. We will tell you plainly whether an engagement makes sense for your firm right now, including when the answer is no.
